Mlscp

Parse and generate MLSCP (Micro LLM Swarm Communication Protocol) commands. Use when communicating with other agents efficiently, parsing compressed commands, or generating token-efficient instructions. Reduces token usage by 70-80% compared to natural language.

MLSCP Skill

MLSCP (Micro LLM Swarm Communication Protocol) is a token-efficient command language for agent-to-agent communication. This skill lets you parse, validate, and generate MLSCP commands without any LLM overhead.

Why Use MLSCP?

Natural Language	MLSCP	Savings
"Please modify the file src/chain_orchestrator.py by adding retry logic at line 47"	F+ s/co > ln47 + 'retry logic'	~75%
"Read the contents of utils/file_manager.py from lines 10 to 50"	F? u/fm > ln10-50	~80%
"Delete the variable 'temp_cache' from config.py"	V- c/c > 'temp_cache'	~70%

Quick Start

Parse a Command

bash

./scripts/mlscp.sh parse "F+ s/co > ln47 + 'retry logic'"

Validate Syntax

bash

./scripts/mlscp.sh validate "F+ s/co > ln47 + 'retry logic'"

Generate Vocabulary

bash

./scripts/mlscp.sh vocab /path/to/project

Compress Natural Language

bash

./scripts/mlscp.sh compress "Add error handling to the main function in app.py"

Command Reference

Operations

Code	Meaning	Example
F+	File add/insert	F+ s/app > ln10 + 'new code'
F~	File modify	F~ s/app > ln10-20 ~ 'updated code'
F-	File delete	F- s/app > ln10-15
F?	File query/read	F? s/app > ln1-100
V+	Variable add	V+ s/app + 'new_var = 42'
V~	Variable modify	V~ s/app > 'old_var' ~ 'new_value'
V-	Variable delete	V- s/app > 'temp_var'
V?	Variable query	V? s/app > 'config_*'

Location Specifiers

• ln47 - Single line
• ln10-50 - Line range
• fn:main - Function name
• cls:MyClass - Class name

Context Blocks

text

CTX{"intent":"resilience","priority":"high","confidence":0.9}

Scripts

• mlscp.sh - Main CLI tool
• vocab.py - Vocabulary generator (Python)

Integration

With Other Agents

When receiving commands from MLSCP-enabled agents:

bash

./scripts/mlscp.sh parse "$INCOMING_COMMAND"

Sending Commands

Generate compact commands for other agents:

bash

./scripts/mlscp.sh compress "Your natural language instruction"

API (Python)

python

from mlscp import parse, MLSCPParser

# Quick parse
cmd = parse("F+ s/co > ln47 + 'retry logic'")
print(cmd.operation)  # OperationType.FILE_ADD
print(cmd.target)     # "s/co"

# With vocabulary
parser = MLSCPParser(vocab_lookup)
cmd = parser.parse("F+ s/co > ln47 + 'code'")
full_path = vocab_lookup.get("s/co")  # "src/chain_orchestrator.py"

Resources

• GitHub: https://github.com/sirkrouph-dev/mlcp
• Grammar Spec: See references/grammar.abnf
• Protocol Definition: See references/protocol.md